Prognostic value of hepatic venous pressure gradient for in-hospital mortality of patients with severe acute alcoholic hepatitis


Dr R. Bañares, Sección de Hepatología, Servicio de Aparato Digestivo, Hospital Gregorio Marañon, Av Dr. Esquerdo 46, 28007 Madrid, Spain.




Hepatic venous pressure gradient (HVPG) has prognostic value in complications and survival of patients with liver cirrhosis. However, the relationship between HVPG and the outcome of acute alcoholic hepatitis (AAH), as well as the specific features of portal hypertension syndrome in this setting, have not been defined.


To evaluate the prognostic value of HVPG and to analyse the degree of portal hypertension and hyperdynamic circulation in patients with severe AAH.


Early measurements of HVPG were performed in 60 patients with severe AAH, and compared with the haemodynamic findings of 37 and 29 liver transplantation candidates with alcoholic or viral end-stage cirrhosis respectively.


Twenty-three patients (38%) died during hospitalization. Portal hypertension and hyperdynamic circulation were more severe in AAH patients. HVPG was greater in non-survivors [26.9 (7.4) vs. 19.4 (5.2) mmHg, P < 0.001]. Only 4/31 (13%) patients with HVPG ≤ 22 mmHg died from the episode of AAH, vs. 19/29 (66%) patients with HVPG > 22 (P < 0.001). Encephalopathy (OR 9.4; CI 1.4–64.8), Model for End-Stage Liver Disease (MELD) score > 25 (OR 7.4; CI 1.4–39.9) and HVPG > 22 mmHg (OR 6.7; CI 1.1–39.9) were independently associated to in-hospital mortality.


Early measurement of HVPG provides important prognostic information on the short-term outcome of patients with severe AAH. In addition, MELD score also seems to be a strong prognostic factor in these patients.


Severe acute alcoholic hepatitis (AAH) is characterized by a severe impairment of liver function frequently associated to clinical manifestations of portal hypertension (PH), with a 40% in-hospital mortality rate.1, 2 The initial factor in the pathophysiology of PH in cirrhosis is the increase in vascular resistance to portal blood flow, not only due to a mechanical consequence of the hepatic histological disorder caused by the liver disease, but also to a dynamic and potentially reversible component secondary to the activation of contractile cells like myofibroblasts and stellate cells. Other major factor contributing to PH is the increase in blood flow through the portal venous system, due to the splanchnic arteriolar vasodilatation and the hyperkinetic circulatory state commonly present in advanced liver disease.3 Although previous reports regarding the specific features of the PH syndrome in AAH have revealed conflicting results,4, 5 a greater degree of PH in these patients has been described, as compared with isolated alcoholic or viral cirrhosis.6 This increase in PH might be possibly due to further activation of the functional component of PH by different vasoactive mediators7–9 which may be increased in alcoholic hepatitis. Furthermore, a recent study has shown that inhibition of tumour necrosis factor-alpha (TNF-α) activity in AAH markedly and persistently reduced hepatic venous pressure gradient (HVPG),10 which closely reflects the degree of PH in both, alcoholic or viral cirrhosis.11, 12

On the other hand, the prognostic value of HVPG has been clearly demonstrated in different clinical situations of chronic liver disease.3, 13 In fact, the presence of oesophageal varices, variceal bleeding or the appearance of complications after surgery for hepatocellular carcinoma has a threshold value of portal pressure;14, 15 in addition, the degree of HVPG reduction after pharmacological treatment predicts the probability of rebleeding, ascites, peritonitis, hepatorenal syndrome, encephalopathy and death.16, 17 Finally, there is evidence that HVPG has prognostic value in complications of cirrhosis and survival of patients with alcoholic liver cirrhosis.18, 19 However, the influence of portal pressure in severe AAH and the value of HVPG measurements in this setting have not been assessed.

The main objective of this study was to evaluate the prognostic value of HVPG in survival of patients with severe AAH. Secondly, we evaluated the degree of PH and hyperdynamic circulation in patients with severe AAH, in comparison to patients with isolated alcoholic or viral cirrhosis.

Patients and methods

The study was performed in all the 60 consecutive patients (42 male, 18 female) admitted at our Liver Unit between 2000 and 2004 with a histologically proven diagnosis of severe AAH. Severity was defined as the existence of a Maddrey’s discriminant function value >32.1 All patients underwent transjugular liver biopsy (TLB) with haemodynamic measurements within 8 days from admission, and were treated with parenteral prednisolone and enteral feeding. In order to compare the haemodynamic findings of AAH patients, haemodynamic data of all patients with decompensated cirrhosis evaluated at our Liver Unit during 2004 were analysed. Those patients under primary or secondary prophylaxis in whom beta-blocker treatment was not withdrawn before the portal pressure measurement were excluded. These patients were being evaluated as liver transplantation candidates, and all of them received systemic and splanchnic haemodynamic measurements within the routine protocol: 37 had advanced alcoholic liver disease without AAH and more than 6 months of alcoholic abstinence, and 29 patients had hepatitis C virus-related end-stage liver cirrhosis.

The study protocol was approved by the local Ethical Committee and signed informed consent was obtained from each patient.

Haemodynamic evaluation

After an overnight fast, a 9 F vascular introducer sheath (Medikit Co. Ltd, Tokyo, Japan) was placed into the right internal jugular vein according to Seldinger’s technique. Then, a Swan–Ganz catheter (Abbott Laboratories, Chicago, IL, USA) was inserted through it into the right pulmonary artery for measurement of pulmonary artery pressure, pulmonary capillary pressure and right atrial pressure (RAP) according to previously described methods.20 Cardiac output (CO) was measured by the thermodilution technique, and systemic vascular resistance (SVR) was calculated as SVR = 80(MAP − RAP)/CO. Afterwards, a 7 F balloon catheter (Cordis SA, Miami, FL, USA) was placed into the right hepatic vein for the assessment of free and wedged hepatic venous pressures (FHVP and WHVP) as previously described.21 The wedged position was confirmed by the absence of reflux after injection of 2 mL of contrast medium, and FHVP was measured with the tip of the catheter <5 cm into the hepatic vein. The zero pressure level was set in the mid-axillary line. The HVPG was calculated as WHVP minus FHVP. The normal value of HVPG ranges from 1 to 5 mmHg. All haemodynamic measurements were recorded and performed at least in duplicate.

Transjugular liver biopsy technique

After haemodynamic measurements, a liver access set with a 7 F sheath and an internal stiffening cannula (William Cook Europe A/S, Bjaeverskov, Denmark) was slid in, placing its tip 4–5 cm inside the right hepatic vein. Subsequently, a 18-gauge and 60-cm-long automated biopsy device (Quick-Core; William Cook Europe A/S), demonstrated as the most effective for obtaining good liver samples in patients with advanced fibrosis and cirrhosis,22 was used. Liver tissue was fixed in 4% formalin for 24 h, sectioned in serial sections of 5-μm thickness, and stained with haematoxylin–eosin, for reticulin, and by Masson’s trichrome stain. AAH was defined as the presence of liver cell necrosis and at least one of the following lesions: Mallory’s hyaline or neutrophilic infiltrate. A single pathologist, without knowledge of clinical or haemodynamic features, estimated liver cell necrosis, neutrophilic infiltrate, Mallory bodies, fatty infiltration and fibrosis by a semiquantitative method (coding 0/1/2/3 for none/slight/moderate/severe degree of occurrence, respectively).

Statistical analysis

Results of all variables are expressed as mean (s.d.) or frequencies. Quantitative variables were compared with Student’s t-test for unpaired data, and discrete variables with the chi-squared test or Fisher exact test, as appropriate, for comparison between in-hospital survivors and non-survivors in the AAH group. The relationship between HVPG and the different histological variables was evaluated by simple correlation analysis.

In order to identify independent predictors for in-hospital mortality, demographic, clinical and haemodynamic variables were included in univariate analysis. Clinically relevant and statistically significant variables (P < 0.1) were then included in a logistic regression model, with the maximum partial likelihood ratio estimate as the criterion for a variable to be entered or removed from the model. In order to obtain correct estimations and to avoid overfitting of the model, several assumptions were made. First, the isolated values of serum bilirubin, creatinine and prothrombin activity were not included in the model because they were already included as components of MELD score. Secondly, only ascites and hepatic encephalopathy from the Child–Pugh classification were included, because the previously described limitations of albumin as a predictive parameter in liver disease23 and because international normalized ratio and bilirubin are included in the MELD score. In order to provide an adequate estimation of the relationship between HVPG and MELD with prognosis over the full range of both variables, we evaluated a first model including both variables as continuous. For providing more clinically useful information, HVPG and MELD score were then included as categorical variables. Selection of the best value of HVPG and MELD score for death prediction was done with receiver operating characteristic (ROC) curves (Figure 1). In order to confirm the changes in haemodynamic parameters in AAH, anova with Bonferroni test for multiple mean comparisons was used to compare haemodynamic parameters among the three groups: AAH, alcoholic cirrhosis without alcoholic hepatitis and HCV-related cirrhosis.

Figure 1.

 ROC curve analyzing the predictive value of HVPG in severe acute alcoholic hepatitis.

All comparisons were two-tailed, and a significance level of 0.05 was used in all statistical tests.


Clinical and laboratory baseline features of the three groups, AAH, alcoholic cirrhosis and viral cirrhosis, are shown in Table 1. Thirty-four (57%) patients with AAH had liver cirrhosis.

Table 1.   Baseline features in the three groups of patients
 AAH (n = 60)AC (n = 37)VC (n = 29)P
  1. Results are expressed as means (s.d.) or absolute frequencies.

  2. a, P < 0.01 for AAH vs. alcoholic cirrhosis; b, P < 0.01 for AAH vs. viral cirrhosis; c, P < 0.001 for AAH vs. alcoholic cirrhosis; d, P < 0.001 for AAH vs. viral cirrhosis.

  3. AAH, acute alcoholic hepatitis; AC, alcoholic cirrhosis; VC, viral cirrhosis; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

Age (years)45 (10)56 (10)55 (13)a,b
Sex (male/female)42/1830/722/7<0.05
Total bilirubin (mg/dL)22 (9)3.2 (1.3)3.1 (2.2)a,b
Albumin (g/dL)3.1 (0.6)3.0 (0.5)2.9 (0.4)0.122
Prothrombin activity (%)42 (12)58 (16)61 (15)a,b
AST (IU/L)115 (63)77 (53)144 (216)0.199
ALT (IU/L)55 (50)59 (44)109 (165)0.135
Creatinine (mg/dL)0.88 (0.64)0.92 (0.21)1 (0.22)0.586
Urea (mg/dL)42 (44)32 (25)36 (11)0.520
Leukocytes (×103/μL)13.3 (7.7)5.4 (2.8)5.3 (2.3)a,b
Ascites (yes/no)59/124/1320/9<0.01
Encephalopathy (yes/no)34/266/318/21<0.01
Variceal bleeding (yes/no)7/534/337/220.50
Child–Pugh score11.1 (1.3)9.3 (1.6)9 (1.8)a,b
MELD score27.4 (8.2)17.2 (1.3)16.1 (0.9)c,d

Haemodynamic findings

Patients with AAH showed features of advanced PH and had a more severe PH syndrome than the alcoholic or viral cirrhotic patients, as demonstrated by a greater HVPG and CO, and a lower SVR (Table 2). No differences in cardiopulmonary pressures among groups were observed.

Table 2.   Haemodynamic findings in the three groups of patients
  1. Results are expressed as means (s.d.).

  2. a, P < 0.05 for AAH vs. alcoholic cirrhosis; b, P < 0.05 for AAH vs. viral cirrhosis; c, P < 0.005 for AAH vs. alcoholic cirrhosis; d, P < 0.005 for AAH vs. viral cirrhosis.

  3. AAH, acute alcoholic hepatitis; AC, alcoholic cirrhosis; VC, viral cirrhosis; HVPG, hepatic venous pressure gradient; CO, cardiac output; MAP, mean arterial pressure; SVR, systemic vascular resistance.

HVPG (mmHg)22.8 (5.5)20.5 (4.6)19.8 (4.8)a,b
CO (L/min)10.1 (2.8)8.1 (2.5)7.5 (2)a,d
MAP (mmHg)79 (14)82 (12.5)77 (11.4)0,45
SVR (dyn s/cm5)621 (189)868 (184)833 (240)c,d

Within the AAH group, HVPG was greater in patients with cirrhosis [22.6 (5.6) vs. 19.2 (3.5) mmHg, P = 0.035]; however, there were no differences regarding CO [10.2 (3) vs. 9.5 (1.4) mmHg, P = 0.598], MAP [77.5 (13.6) vs. 84.9 (14.8) mmHg, P = 0.187] or SVR [602 (196) vs. 696 (144) mmHg, P = 0.248].

Interestingly, HVPG was also significantly greater in patients with renal failure, as compared with those patients with normal renal function [25.1 (8.1) vs. 22.3 (5.5) mmHg, P = 0.008].

Five patients from the AAH group accepted a second haemodynamic evaluation and TLB after 11 (4) months, all of whom had maintained alcoholic abstinence. HVPG and CO markedly decreased in these patients [from 23.7 (3.9) to 9 (0.9) mmHg, P = 0.038 and from 12 (4.2) to 9.9 (2.6) L/min, P = 0.046], despite the fact that all cases showed features of established cirrhosis in both liver biopsies.

Histological variables and haemodynamic findings

The degree of liver cell necrosis and architectural distortion had a weak correlation with the severity of PH (r = 0.416, P = 0.016; r = 0.484, P = 0.031, respectively), whereas no correlations were found between HVPG and degree of inflammation, fatty change or occurrence of Mallory bodies.

Survival analysis

Twenty-three patients (38%) with AAH died during hospitalization. Six of 23 deaths were clearly related to acute variceal bleeding; in the remaining cases, a single cause of death was not identified because all patients had several concomitant life-threatening conditions. A 22 mmHg of HVPG value and a 25 point of MELD score were determined as the best cut-off values for prediction of death. Age, MELD score, HVPG, and occurrence of encephalopathy, bacterial infection or renal failure were significantly related to in-hospital death in univariate analysis (Table 3).

Table 3.   Variables associated to mortality in univariate analysis
 In-hospital deathsAliveP
  1. Results are expressed as means (s.d.) and proportions.

  2. Renal failure was defined as a Cr value > 1.5 mg/dL.

  3. n, number; HVPG, hepatic venous pressure gradient; CO, cardiac output; MAP, mean arterial pressure; SVR, systemic vascular resistance.

Age (years)48 ( 9.8)42 (8.9)0.019
Sex [n (%)]
 male15 (35.7)27 (64.3)0.527
 female8 (44.4)10 (55.6)
HVPG (mmHg)26.9 (7.4)19.4 (5.2)<0.001
HVPG > 22  mmHg [n (%)]
 No4 (12.9)27 (87.1)<0.001
 Yes19 (65.5)10 (34.5)
CO (L/min)10.9 (3.1)9.5 (2.4)0.122
MAP (mmHg)76.7 (11.2)82.1 (16.6)0.230
SVR (dyn s/cm5)562 (184)660 (185)0.134
Total bilirubin (mg/dL)27.2 (8.4)20.3 (8.1)0.002
MELD score34.7 (7.5)22.9 (5.8)<0.001
MELD > 25 [n (%)]
 No3 (10)27 (90)<0.001
 Yes20 (66.7)10 (33.3)
Child–Pugh [n (%)]
 B1 (12.5)7 (87.5)0.138
 C22 (42.3)30 (57.7)
Ascites [n (%)]
 No01 (100)0.430
 Yes23 (39)36 (61)
Encephalopathy [n (%)]
 No2 (7.7)24 (92.3)<0.001
 Yes21 (61.8)13 (38.2)
Bacterial infection [n (%)]
 No3 (13.6)19 (86.4)0.003
 Yes20 (52.6)18 (47.4)
Renal failure [n (%)]
 No4 (11.4)31 (88.6)<0.001
 Yes19 (76)6 (24)
Variceal bleeding [n (%)]
 No17 (34)33 (66)0.126
 Yes6 (60)4 (40)

In multivariate analysis, only HVPG, MELD score and encephalopathy were independent predictors of in-hospital mortality (Table 4).

Table 4.   Independent prognostic factors for in-hospital mortality in logistic regression analysis
 Odds ratio95% confidence interval
  1. Model 1: HVPG and MELD score are introduced as continuous variables. Model 2: HVPG and MELD score are introduced as categorical variables.

Model 1
 MELD score1.181.04–1.34
Model 2
 MELD score > 257.41.4–39.9
 HVPG > 226.71.1–39.9

In order to evaluate the possible influence of the presence of cirrhosis on AAH patients, we have separately analysed data from patients with and without cirrhosis. No differences were found regarding the degree of liver dysfunction (as estimated by means of MELD score and Child–Pugh classification), mortality and haemodynamic variables other than HVPG (Table 5).

Table 5.   Comparative analysis of cirrhotic and non-cirrhotic patients. Results are expressed as means (s.d.) and proportions. Renal failure was defined as a Cr value > 1.5 mg/dL
VariableAAH with cirrhosisAAH without cirrhosisP
  1. AAH, acute alcoholic hepatitis; HVPG, hepatic venous pressure gradient; CO, cardiac output; MAP, mean arterial pressure; SVR, systemic vascular resistance; INR, international normalized ratio; n, number.

HVPG (mmHg)22.6 (5.6)19.2 (3.5)0.035
CO (L/min)10.2 (3)9.5 (1.4)0.598
MAP (mmHg)77.5 (13.6)84.9 (14.8)0.187
SVR (dyn s/cm5)602 (196)696 (144)0.248
MELD score28.2 (8.8)26.5 (8.5)0.454
Bilirubin (mg/dL)22.9 (9.6)22.9 (7.9)0.518
Albumin (g/dL)2.7 (0.4)2.4 (2.1)0.434
INR2.7 (1.1)2.4 (0.6)0.301
Creatinine (mg/dL)1.2 (0.7)1.1 (0.7)0.607
Platelets (×103/μL)106.5 (79)123.9 (87)0.419
Death [n (%)]14 (41.2)9 (34.6)0.604
Renal failure [n (%)]16 (47.1)9 (34.6)0.333
Infection [n (%)]22 (64.7)16 (61.5)0.801
Child–Pugh [n (%)]
 B4 (11.8)4 (15.4)0.683
 C30 (88.2)22 (84.6)
Sex [n (%)]
 Female11 (32.4)7 (26.9)0.649
 Male23 (67.6)19 (73.1)


One of the most important contributions of HVPG measurement in the clinical setting is its prognostic value in different situations of cirrhotic patients. In this study, we show for the first time that the degree of portal pressure may influence prognosis in AAH patients. In fact, HVPG has an independent predictive value to determine in-hospital mortality in patients with severe AAH. Indeed, besides well-known predictors of short-term prognosis in chronic liver disease such as MELD score and encephalopathy, a HVPG over 22 mmHg was a strong independent predictor for mortality. Our finding is clinically relevant as severe AAH has a high in-hospital mortality rate (around 40%), and identification of early and accurate predictors of short-term outcome of these patients is needed. Recent studies have suggested several prognostic indicators in AAH related to pathophysiological events,24 clinical evolution after treatment25 and indirect evaluation of PH.26 In this context, HVPG measurement has a good profile to become a predictor of short-term outcome, as it is easily performed during TLB, which is useful in order to establish the gold-standard histological diagnosis of AAH. The 22 mmHg threshold value clearly determines two different populations with a 13% (4/31) vs. 66% (19/29) mortality rate between those who had HVPG below or above this value. Therefore, a more aggressive therapeutic approach may be warranted according to HVPG value. It is important to emphasize that the inclusion of HVPG as a continuous variable provides an adequate estimation of risk over the full range of the variable, indicating a clear association between HVPG and clinical outcome. These data provide further insight about the prognostic value of HVPG that has been identified in other clinical settings of end-stage liver disease.9, 13, 17

Additionally, our results provide more information about the prognostic value for in-hospital mortality of the MELD score in patients with severe AHH. A recent study suggests that MELD score might be a more precise test than Maddrey’s discriminant function (DF) and Child–Turcotte–Pugh classification for the prediction of in-hospital mortality in these patients.27 In fact, admission, first week, and first week change in MELD score are significantly independent predictors. Another study has recently shown that MELD score is as useful as DF for predicting 30-day and 90-day mortality in patients with AAH, with the cut-off value of 21 as the most accurate one.28 However, not all the patients included in this study had histological confirmation and DF > 32, and these factors might account for the different cut-off point with regard to our study.

Although MELD score is a more easily applicable test than HVPG measurement, these patients frequently need to be transferred to specialized Liver Units in tertiary referral centres, where TLB is necessary to ascertain a definite diagnosis of AAH. Taking into account its marked prognostic value, we suggest that HVPG measurements should be done in the setting of TLB.

Our model may have some limitations. First, the relatively low number of events (23 deaths in 60 patients) may induce overfitting of the model if an excessive number of variables is included in the analysis. However, the strategy in the definition of variables to be introduced in the model allows us to build a final model with eight events for each finally retained variable, that is clearly acceptable.29 Another possible limitation is that we have not included in the model the early change in bilirubin level that has been suggested as an accurate prognostic marker in AAH.25 However, the inclusion in the model of MELD score precludes the use of early change in bilirubin level due to colinearity.

Another interesting finding of our study is the existence of a greater HVPG and a more pronounced hyperdynamic circulation in patients with AAH than in those with end-stage liver cirrhosis. The principal mechanism which may lead to that worsening of PH syndrome may be related to the characteristic inflammatory profile of AAH. In fact, an increase of TNF-α activity has been reported in patients with AAH, especially during clinically stressful conditions. The influence of this cytokine on PH is clearly shown by the existence of increased levels of circulating TNF-α in the setting of in vivo PH30, 31, as well as by the capacity of some drugs with anti-TNF-α activity (like thalidomide or anti-TNF-α antibodies) to decrease hyperdynamic circulation and portal pressure.32, 33 Moreover, a recent study has reported a marked and long-lasting improvement in splanchnic and systemic haemodynamic parameters of patients with severe AAH following treatment with anti-TNF-α antibodies.10 In addition, splanchnic vasodilation may produce a homeostatic response with hyperactivation of neuro-hormonal systems involved in the modulation of intrahepatic vascular tone, as demonstrated by studies in perfused cirrhotic livers.7 This fact is supported by a recent study that evaluated the effects of albumin dialysis (Molecular Adsorbents Recirculating System; MARS) in patients with acute on chronic liver failure, mainly due to severe AAH. A significant improvement in HVPG, CO, MAP and SVR was observed after therapy. Furthermore, plasma renin activity and serum norepinephrine significantly decreased, and this reduction was correlated to the reduction in HVPG.34 The marked improvement of HVPG that we observed in those five patients, who had a second haemodynamic study, even in the presence of established cirrhosis, is clearly in agreement with the existence of a functional, and therefore reversible, component of PH. It has been previously reported4, 6 that the higher portal pressure observed in AAH patients may be partially related to the degree of liver cell necrosis. In our study, we observed a weak correlation between this histological finding and HVPG. However, it should be cautiously interpreted, because reproducibility of grading necrosis in those patients is not completely established. The precise mechanism by which hepatocellular necrosis could raise portal pressure is not fully understood: it might be due to an increase in hepatocyte volume.35, 36 On the other hand, the increment of portal pressure in fulminant hepatitis has been associated to massive hepatocellular necrosis, which collapses the sinusoidal frame.37, 38

In conclusion, our findings suggest the strong influence of HVPG in the short-term prognosis of severe AAH, which supports its incorporation to the everyday management of severe acute alcoholic patients, although this point should be evaluated in specifically designed trials. In addition, PH syndrome is more severe in AAH than in end-stage alcoholic or viral cirrhosis. Finally, MELD score also seems to be a strong prognostic factor in these patients.


Declaration of personal interests: CR received financial support from Fondo de Investigaciones Sanitarias (FIS) CM 03/00037. OL and JG-C received financial support from the Fundacion de Investigacion Biomedica del Hospital Gregorio Marañon, Madrid, Spain.

Declaration of funding interests: This work was supported in part by CIBEREHD, Instituto de Salud Carlos III, Spain.